Enforcing access restrictions for fine-tuning machine learning models

Machine-learned models and data-driven systems have been increasingly used to help make decisions in various application domains. These applications have provided benefits such as improved accuracy, increased productivity, and cost savings. This trend is the result of a confluence of factors, such as ubiquitous connectivity, the ability to collect, aggregate, and process large amounts of fine-grained data using cloud computing, and improved access to increasingly sophisticated machine learning models that can analyze this data.

While embodiments are described herein by way of example for several embodiments and illustrative drawings, those skilled in the art will recognize that embodiments are not limited to the embodiments or drawings described. It should be understood, that the drawings and detailed description thereto are not intended to limit embodiments to the particular form disclosed, but on the contrary, the intention is to cover all modifications, equivalents and alternatives falling within the spirit and scope as described by the appended claims. The headings used herein are for organizational purposes only and are not meant to be used to limit the scope of the description or the claims. As used throughout this application, the word “may” is used in a permissive sense (e.g., meaning having the potential to), rather than the mandatory sense (e.g., meaning must). Similarly, the words “include,” “including,” and “includes” mean including, but not limited to.

It will also be understood that, although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first contact could be termed a second contact, and, similarly, a second contact could be termed a first contact, without departing from the scope of the present invention. The first contact and the second contact are both contacts, but they are not the same contact.

Various techniques of enforcing access restrictions for fine-tuning machine learning models are described herein. Machine learning refers to a discipline by which computer systems can be trained to recognize patterns through repeated exposure to training data. In unsupervised learning, a self-organizing algorithm learns previously unknown patterns in a data set without any provided labels. In supervised learning, this training data includes an input that is labeled (either automatically, or by a human annotator) with a “ground truth” of the output that corresponds to the input. A portion of the training data set is typically held out of the training process for purposes of evaluating/validating performance of the trained model. The use of a trained model in production is often referred to as “inference,” or a “prediction” during which the model receives new data that was not in its training data set and provides an output based on its learned parameters.

Fine-tuning may refer to techniques to adapt the features of a previously trained machine learning model (e.g., the weights) according to additional training data that may “tune” or otherwise adapt the trained machine learning model's performance to specific uses or scenarios included in the additional training data. For example, a computer vision model that performs object classification generally may be tuned to recognize a particular category of objects, such as traffic signs, in image data. However, there may be scenarios where fine-tuning of a trained machine learning model is desirable, but modification of the trained machine learning model itself is not supported or allowed due to access restrictions.

For example, some machine learning models are developed as the result of significant technological effort and resource costs. Appropriate data sets may have to be curated and the architecture of the machine learning model designed to provide a high-performing machine learning model. Some of these machine learning models can be extremely large using, for instance, billions of parameters, allowing the model to be adaptable to a wide category of use cases and tasks, such as text and image generation and summarization. These machine learning models, which are sometimes referred to as “foundation models”, may perform well without any adaptation. However, in many scenarios, better performance can be achieved if the models are fine-tuned to specific uses cases. Given the technological efforts and resource costs expended to develop and train these machine learning models, model providers may impose access restrictions on the content of the models (e.g., the weights of model parameters), as it would otherwise have to surrender proprietary model information if the content of the models were to be accessible. Accordingly, techniques for fine-tuning machine learning models that preserve the access restrictions of model providers may be highly desirable.

Consumers of these machine learning models may not be without their own data privacy concerns. Training data used to fine-tune machine learning models may be subject to regulatory, proprietary, sensitive, or other access restrictions that would prohibit or make it undesirable to grant access to the training data, especially to model providers which could make use of the training data to enhance the performance of the machine learning model more generally. Accordingly, techniques for fine-tuning machine learning models that preserve the access restrictions of model consumers may also be highly desirable.

In order to satisfy the access restrictions of both model providers and model consumers, a machine learning service can support fine-tuning techniques that make the restricted data opaque to entities without access (e.g., a model provider cannot access model consumer tuning data and the resulting fine-tuned model, and a model consumer cannot access the trained machine learning model directly). By implementing these fine-tuning techniques, the performance advantages offered by some machine learning models, such as foundation models, can be made available and adaptable to specific use cases through fine-tuning techniques, improving the performance of systems, services, or applications that utilize fine-tuned machine learning models without violating the access restrictions of both model providers and model consumers.

illustrates a logical block diagram of enforcing access restrictions for fine-tuning machine learning models, according to some embodiments. Model providermay develop and train a machine learning model that may be made available to various model consumers, such as model consumer. Machine learning service, which may be a machine learning service of a services provider (e.g., a cloud computing services provider) similar to that of machine learning servicediscussed in detail below with regard to, or implemented differently (e.g., as a standalone service), may support enforcing access restrictions for the trained machine learning model offered by model provider. Provider access restrictionsmay limit access to the trained machine learning model, such as by prohibiting access to model consumers, like model consumer.

Model consumermay have identified the trained machine learning model for incorporation into a system, service, or application. In order to adapt performance of the trained machine learning model to a desired use case, model consumermay initiate model fine-tuningat machine learning service. As part of fine-tuning, model consumermay provide a tuning data set, as indicated at. The tuning data set may be subject to consumer access restrictions, which may, for example, limit access to the tuning data set, as well as any resulting tuned weights, such as by prohibiting access to model provider. In some embodiments, access restrictions may be described as preventing exfiltration of protected data to a different entity (e.g., a model provider cannot access model consumer tuning data and the resulting fine-tuned model weights, and a model consumer cannot access the weights of a trained machine learning model directly).

Model fine-tuningmay perform various fine-tuning techniques that do not violate the consumer access restrictionsand provider access restrictions, such as the various fine-tuning techniques discussed below with regard to, that generate additional tuned sets of weights (different from the trained model weights provided at). Once generated, the set of tuned weights may be provided to model consumerfor various uses, including deployment as discussed in detail below with regard to. As indicated at, the set of tuned weights may not be useable without the trained model weights. Instead, the set of tuned weights may be used in combination with the trained model weights to implement the fine-tuned version of the machine learning model. For example, both the tuned weights and the trained model weights may be used when generating an inference using the fine-tuned model.

Please note that the previous description is a logical illustration of a machine learning service and thus is not to be construed as limiting as to other embodiments of a machine learning service.

This specification continues with a general description of a provider network that implements multiple different services, including a machine learning service, which may implement enforcing access restrictions for fine-tuning machine learning models, according to some embodiments. Then various examples of, including different components/modules, or arrangements of components/module that may implement enforcing access restrictions for fine-tuning machine learning models are discussed. A number of different methods and techniques to implement enforcing access restrictions for fine-tuning machine learning models are then discussed, some of which are illustrated in accompanying flowcharts. Finally, a description of an example computing system upon which the various components, modules, systems, devices, and/or nodes may be implemented is provided. Various examples are provided throughout the specification.

illustrates an example provider network that may implement a machine learning service that implements enforcing access restrictions for fine-tuning machine learning models, according to some embodiments. Service(s) providermay be a private or closed system or may be set up by an entity such as a company or a public sector organization to provide one or more services (such as various types of cloud-based storage) accessible via the Internet and/or other networks to clients, in one embodiment. Service(s) providermay be implemented in a single location or may include numerous data centers hosting various resource pools, such as collections of physical and/or virtualized computer servers, storage devices, networking equipment and the like (e.g., computing systemdescribed below with regard to), needed to implement and distribute the infrastructure and services offered by the service(s) provider, in one embodiment. In some embodiments, service(s) providermay implement various computing resources or services, such as machine learning service, storage service(s), and/or any other type of network-based services(which may include a virtual compute service and various other types of storage, database or data processing, analysis, communication, event handling, visualization, data cataloging, data ingestion (e.g., ETL), and security services), in some embodiments.

In various embodiments, the components illustrated inmay be implemented directly within computer hardware, as instructions directly or indirectly executable by computer hardware (e.g., a microprocessor or computer system), or using a combination of these techniques. For example, the components ofmay be implemented by a system that includes a number of computing nodes (or simply, nodes), each of which may be similar to the computer system embodiment illustrated inand described below, in one embodiment. In various embodiments, the functionality of a given system or service component (e.g., a component of machine learning servicemay be implemented by a particular node or may be distributed across several nodes. In some embodiments, a given node may implement the functionality of more than one service system component (e.g., more than one data store component).

Machine learning servicemay implement interfaceto allow clients (e.g., client(s)or clients implemented internally within service(s) provider, such as a client application hosted on another provider network service like an event driven code execution service or virtual compute service) to compress, train, and deploy machine learning models (e.g., neural networks), or various other machine learning model development, deployment, or analysis features. For example, machine learning servicemay implement interface(e.g., a graphical user interface, programmatic interface that implements Application Program Interfaces (APIs) and/or a command line interface) may be implemented so that a client can submit, edit, or otherwise implement various different model development, deployment, labeling or other management requests. For example, interfacemay include development and deployment environment, which may provide a training script or other code editor with various development tools to create, submit, and/or monitor machine learning pipeline with a tuning job and/or create a model package, as discussed below.

Machine learning servicemay implement a control planeto perform various control operations to implement the features of machine learning service. For example, control plane may monitor the health and performance of requests at different components, such as training as part of model development, execution of machine learning models as part of model deploymentor development and training of machine learning models as part of model development. For example, if a node or other component fails, a request fails, or other interruption occurs, control planemay be able to restart a job to complete a request (e.g., instead of sending a failure response to the client). Control planemay, in some embodiments, may arbitrate, balance, select, or dispatch requests to different node(s), in various embodiments. For example, control planemay receive requests interfacewhich may be a programmatic interface, and identify an available node to begin work on the request.

Machine learning servicemay implement model developmentto develop, configure, program, define, and/or otherwise execute training jobs on various machine learning models using data sets, such as data setsin storage servicesacross one or more training nodes (which may include one or more respective processing devices for training, such as GPUs). In some embodiments machine learning servicemay offer various virtual machines, instances, containers, images, or other applications on these training nodes that may implement various machine learning training frameworks (e.g., TensorFlow, PyTorch, MXNet, and XGBoost, among others) upon which machine learning models may be specified or otherwise described using, for instance, a development environment, and executed. Various tests or other development operations for machine learning models may also be performed. In some embodiments, the various files, configuration information, and other data for machine learning model development may be organized as a project (or other collection) and stored, versioned, or otherwise managed by model development(e.g., as a collection of one or more files or data objects in storage services, including data setsand ML models). Training jobs may be submitted to training nodes (e.g., via development environment or other interfaces) to train machine learning models on identified data set(s). As discussed in detail below with regard to, opaque fine-tuning requests may be submitted as well.

In various embodiments, machine learning servicemay implement model deployment, which may deploy a trained machine learning model on resources (e.g., virtual compute instances or containers) to receive and return inferences or other results according to requests or other inputs to the deployed model. For example, different types or configurations of resources (e.g., virtual compute instances with various hardware capabilities, including different amounts of processing capacity, memory, storage, and/or specialized hardware, such as GPUs and tensor processor units (TPUs)) may be provisioned or otherwise obtained from other services of service(s) provider(e.g., a virtual compute service) and then the machine learning model deployed to that provisioned resource along with various software or other applications to support the receipt of requests for inferences and return inferences. As discussed in detail below with regard to, opaque fine-tuned machine learning models may be deployed, in some embodiments.

Data storage service(s)may implement different types of data stores for storing, accessing, and managing data on behalf of clientsas a network-based service that enables clientsto operate a data storage system in a cloud or network computing environment. Data storage service(s)may also include various kinds relational or non-relational databases, in some embodiments, data storage service(s)may include object or file data stores for putting, updating, and getting data objects or files, in some embodiments. For example, one data storage servicemay be an object-based data store that allows for different data objects of different formats or types of data, such as structured data (e.g., database data stored in different database schemas), unstructured data (e.g., different types of documents or media content), or semi-structured data (e.g., different log files, human-readable data in different formats like JavaScript Object Notation (JSON) or Extensible Markup Language (XML)) to be stored and managed according to a key value or other unique identifier that identifies the object. In at least some embodiments, data storage service(s)may be treated as a data lake. For example, an organization may generate many different kinds of data, stored in one or multiple collections of data objects in a data storage service. The data objects in the collection may include related or homogenous data objects, such as database partitions of sales data, as well as unrelated or heterogeneous data objects, such as image data files (e.g., digital photos or video files) audio files and web site log files. Data storage service(s)may be accessed via programmatic interfaces (e.g., APIs) or graphical user interfaces.

Generally speaking, clientsmay encompass any type of client that can submit network-based requests to service(s) providervia network, including requests for machine learning service(e.g., a request to create or perform an explanation job, interact with development and management environment, etc.). For example, a given clientmay include a suitable version of a web browser, or may include a plug-in module or other type of code module that can execute as an extension to or within an execution environment provided by a web browser. In some embodiments, such an application may include sufficient protocol support (e.g., for a suitable version of Hypertext Transfer Protocol (HTTP)) for generating and processing network-based services requests without necessarily implementing full browser support for all types of network-based data. That is, clientmay be an application that can interact directly with service(s) provider. In some embodiments, clientmay generate network-based services requests according to a Representational State Transfer (REST)-style network-based services architecture, a document- or message-based network-based services architecture, or another suitable network-based services architecture.

In some embodiments, a clientmay provide access to service(s) providerto other applications in a manner that is transparent to those applications. Clientsmay convey network-based services requests (e.g., access requests to configure or perform explanation jobs) via network, in one embodiment. In various embodiments, networkmay encompass any suitable combination of networking hardware and protocols necessary to establish network-based-based communications between clientsand service(s) provider. For example, networkmay generally encompass the various telecommunications networks and service providers that collectively implement the Internet. Networkmay also include private networks such as local area networks (LANs) or wide area networks (WANs) as well as public or private wireless networks, in one embodiment. For example, both a given clientand service(s) providermay be respectively provisioned within enterprises having their own internal networks. In such an embodiment, networkmay include the hardware (e.g., modems, routers, switches, load balancers, proxy servers, etc.) and software (e.g., protocol stacks, accounting software, firewall/security software, etc.) necessary to establish a networking link between given clientand the Internet as well as between the Internet and service(s) provider. It is noted that in some embodiments, clientsmay communicate with service(s) providerusing a private network rather than the public Internet.

As discussed above with regard to, model providers may provide trained machine learning models for model consumers to use as part of an application. In at least some embodiments, machine learning model catalogmay be implemented as part of service(s) providerto provide for the submission of new trained machine learning modelsas well as allowing for models to be used through model subscription management.is a logical block diagram illustrating interactions with a machine learning model catalog, according to some embodiments. As indicated at, a request to submit a trained machine learning model may be made to model submission.

Model submissionmay apply various validation, formatting, encryption, or other access management techniques and store the trained machine learning model in a service accountin storage services, as indicated at. Service accountmay act as an escrow account on behalf of a model provider, storing trained ML modeland training/inference instructions, such that access is restricted unless granted by the model provider (e.g., for use as opaque fine-tuning as discussed in detail below). In some embodiments, as part of model submission, both the trained ML model(e.g., the architecture and parameter weights for a neural network) and the training/inference instructionsmay be provided (e.g., as scripts, executables, images, etc.). As part of submission request, various access restrictions may be specified in including whether (if any) opaque fine-tuning and/or non-opaque fine-tuning may be performed on the trained machine learning model. For example, some submitted trained machine learning models may be used (e.g., invoked to perform inference requests) without allowing for any fine-tuning in accordance with the submission request.

As indicated at, a request to subscribe to a trained machine learning model may be received at model subscription management. For instance, various indexes or other data structures may be made searchable via an interface of ML model catalog. In some embodiments, the search request may include features to filter by supported fine-tuning type (e.g., opaque fine-tuning, non-opaque fine-tuning). Various descriptive information of many different machine learning models submitted and offered through ML model catalogmay be provided, including whether opaque fine-tuning is supported (or non-opaque fine-tuning is supported) for a machine learning model. For subscription requests, a registry or other subscriber index may be updated to include a model consumer (e.g., an account) for a requested machine learning model. When it is time to train or deploy the machine learning model, model subscription managementmay verify a subscription and provide access credentials to obtain the ML model and training/inference instructions, as indicated at.

is as logical block diagram illustrating interactions to request opaque fine-tuning, according to some embodiments. As indicated at, a request for opaque fine-tuning for trained ML model may be received at model development. In some embodiments, other terminology such as privacy-preserving, restricted fine-tuning, or proprietary fine-tuning, among other terms which may invoke fine-tuning techniques that enforce both model provider and model consumer access restrictions on their respective data. The requestmay include specified hyperparameters or other configuration to apply to the training instructions (e.g., including overriding some default hyperparameters). Requestmay specify one of many different supported fine-tuning techniques (as discussed below with regard to). Model developmentmay dispatch the opaque fine-tuning jobto provisioned computing resources. A network configuration that enforces access restrictionsmay be implemented (e.g., using virtual private network techniques to prevent exfiltration). For example, specialized network endpoints that impose firewalls other network traffic management techniques can be implemented that prevent outbound data from provisioned computing resources, allowing a tuning process to get weights/training instructionsto perform without sending data to external destinations outside of(e.g., without allowing tuning datato be sent to other destinations by provisioned computing resources and trained ML model weights/training instructionsto be sent to other destinations).

Provisioned computing resources may getthe weights and training instructions from service account, storing trained ML model weightsand training instructions. Provisioned computing resourcesmay also gettuning data from consumer accountstoring tuning data. Provisioned computing resourcesmay then perform opaque fine-tuning techniques, as discussed in detail below with regard to, and store the tuned weights, as tuned weightsin consumer account.

is a logical block diagram illustrating techniques for opaquely fine-tuning a machine learning model, according to some embodiments. Opaque fine-tuning techniques may be performed such that the underlying weights of the trained machine learning model do not have to be modified in order to modify performance of the model overall, allowing for fine-tuning without direct model modification. Instead, as depicted in, given some input featuresthat are provided to both trained ML model weightsand fine-tuned ML model weights, the resulting output can be combined (e.g., concatenated) and output(e.g., as an embedding which can be decoded to provide an inference).

Various different types of opaque fine-tuning techniques may include Parameter Efficient Fine-Tuning (PEFT) techniques, in some embodiments. Parameter efficient fine-tuning refers to a set of fine-tuning techniques that do not require updating all the model weights. Instead, just a subset of the weights are updated. A notable component of PEFT methods only fine-tune a small number of (extra) model parameters. The following are some examples of PEFT techniques.

LoRA: Low Rank Adaptation is a technique where the pre-trained weights from the provided machine learning model are frozen and a smaller set of incremental weights are trained using the tuning data set. During inference, the results of the incremental weights are added to the frozen ones. LoRA can yield better results than incremental fine-tuning and be faster to fine-tune.

AdaLoRA: LoRA but with an adaptive learning rate that adjusts based on the curvature information of the loss landscape.

Prefix Tuning: The idea behind prefix-tuning is to optimize a continuous vector that is prepended to the input of a language model. This vector, also known as a “prefix”, is used to guide the model's generation process. Prefix-tuning only adjusts the prefix, leaving the rest of the model parameters fixed.

P-Tuning: A set of trainable parameters (P) as additional tokens are introduced at the beginning of the input sequence. These parameters are learned during the fine-tuning process and are task-specific.

Prompt Tuning: A mechanism for learning “soft prompts” to condition frozen language models to perform specific downstream tasks from labeled examples.

RLHF: Leveraging reinforcement learning to “teach” a model with a reward model tuned on human feedback data.

is a logical block diagram illustrating interactions to deploy an opaquely fine-tuned machine learning model, according to some embodiments. As indicated at, a request to deploy an opaque fine-tuned ML model may be received. The request may include various information to configure deployment including a location (e.g., provider network region) and host system configuration (e.g., computing resource type, such as an instance type). In some embodiments, a model package (e.g., a configuration file or other description of the originally trained and tuned model weights along with the training and inference instructions may be first created (e.g., by model deploymentin response to a request), which may then be referenced in request(e.g., causing model deploymentto obtain the model deployment package from consumer account(not illustrated) to provide to provisioned computing resources). Model deploymentmay place the opaque fine-tuned ML model on provisioned computing resources. A network configuration that enforces access restrictionsmay be implemented (e.g., using virtual private network techniques to prevent exfiltration). For instance, although provisioned computing resource(s)may be requested and charged/billed to a user account that submitted deployment request, network configurationmay prevent any outbound data that includes the trained ML model weightsand inference instructions.

Provisioned computing resources may getweights and inference instructions from service account, storing trained ML model weightsand inference instructions. Provisioned computing resourcesmay gettuned weightsfrom consumer account. Using the combination of tuned weightsand trained ML model weights, inferences may be performed when request, as indicated atand returned, as indicated at. For example, a client application, hosted in other services of service(s) providerassociated with the same account that subscribed to the ML model and tuned the ML model at, may generated and send inference requestand perform various operation(s) based on inferencereturned to that client application.

Althoughhave been described and illustrated in the context of a provider network implementing a machine learning service, the various components illustrated and described inmay be easily applied to other machine learning systems. As such,are not intended to be limiting as to other embodiments.

is a high-level flowchart illustrating various methods and techniques for enforcing access restrictions for fine-tuning machine learning models, according to some embodiments. As indicated at, a request to perform fine-tuning on a trained machine learning model using a specified tuning data set via an interface of a machine learning service may be received, in some embodiments. The request may specify a fine-tuning type, in some embodiments, that specifies access-restriction enforced fine-tuning for a model consumer and model provider. As discussed above with regard to, various hyperparameters or other configuration for fine-tuning may be allowed, in some embodiments. For instance, a model provider may support the selection by a consumer of different hyperparameters, which may be included in the request.

As indicated at, a set of weights of the trained machine learning model, the tuning data set, and tuning instructions for performing the fine-tuning on the trained machine learning model, where the trained machine learning model is obtained from a model provider with provider access restrictions on the set of weights, and the tuning data set is obtained from a model consumer with consumer access restrictions on the tuning data set, in some embodiments. For example, escrow or other access-restricted data storage locations may be accessed by the machine learning service without providing access to a consumer that submitted the request to perform fine-tuning.

As indicated at, the tuning instructions may be executed to generate a set of fine-tuned weights at computing resource(s) of the machine learning service provisioned to perform the fine-tuning us the tuning data set and the set of weights of the trained machine learning model, where the fine-tuned weights are used in combination with the set of weights of the trained machine learning model to perform inferences, and where the execution enforces the provider and consumer access restrictions, in some embodiments. As discussed above with regard to, various fine-tuning techniques that utilize both the set of weights of the trained model and new weights to be used as the fine-tuned weights may be performed (e.g., PEFT techniques such as LORA, AdaLORA, and Prefix Tuning among others). In some embodiments, fine-tuned weights may be referred to as the “delta weights”.

As indicated at, the set of fine-tuned weights in a storage location that satisfies the consumer access restrictions may be stored, in some embodiments. For example, access to the storage location may be limited to requests verified as authorized by an account that requested fine-tuning and provided the tuning data set.

Once fine-tuned, the machine learning model can be deployed on behalf of the model consumer for various uses.is a high-level flowchart illustrating various methods and techniques for deploying a fine-tuned machine learning model and enforcing access restrictions, according to some embodiments. As indicated at, a request to deploy a fine-tuned machine learning model, may be received, in some embodiments. As noted above, various deployment configuration information, including host system resources or capabilities and host system location may be specified.

As indicated at, computing resource(s) may be provisioned to host the fine-tuned machine learning model, in some embodiments. For example, the specified deployment configuration information may be used to obtain, prepare and/or configure a host system for the fine-tuned machine learning model. Various network configuration operations may be performed to prepare network resources (e.g., firewalls, specialized network endpoints, virtual private networks, etc.) to enforce access restrictions.

As indicated at, a set of weights of the trained machine learning model, a tuned set of weights, and inference instructions for performing inferences on the fine-tuned machine learning model may be obtained, in some embodiments. As indicated at, the computing resources may be made available to perform inferences according to the inference instructions using a combination of the set of weights of the trained machine learning model and the tuned set of weights while enforcing model provider access restrictions and consumer access restrictions, in some embodiments.

The methods described herein may in various embodiments be implemented by any combination of hardware and software. For example, in one embodiment, the methods may be implemented on or across one or more computer systems (e.g., a computer system as in) that includes one or more processors executing program instructions stored on one or more computer-readable storage media coupled to the processors. The program instructions may implement the functionality described herein (e.g., the functionality of various servers and other components that implement the network-based virtual computing resource provider described herein). The various methods as illustrated in the figures and described herein represent example embodiments of methods. The order of any method may be changed, and various elements may be added, reordered, combined, omitted, modified, etc.

Embodiments of enforcing access restrictions for fine-tuning machine learning models as described herein may be executed on one or more computer systems, which may interact with various other devices. One such computer system is illustrated by. In different embodiments, computer systemmay be any of various types of devices, including, but not limited to, a personal computer system, desktop computer, laptop, notebook, or netbook computer, mainframe computer system, handheld computer, workstation, network computer, a camera, a set top box, a mobile device, a consumer device, video game console, handheld video game device, application server, storage device, a peripheral device such as a switch, modem, router, or in general any type of computing device, computing node, compute node, or electronic device.

In the illustrated embodiment, computer systemincludes one or more processorscoupled to a system memoryvia an input/output (I/O) interface. Computer systemfurther includes a network interfacecoupled to I/O interface, and one or more input/output devices, such as cursor control device, keyboard, and display(s). Display(s)may include standard computer monitor(s) and/or other display systems, technologies or devices. In at least some implementations, the input/output devicesmay also include a touch- or multi-touch enabled device such as a pad or tablet via which a user enters input via a stylus-type device and/or one or more digits. In some embodiments, it is contemplated that embodiments may be implemented using a single instance of computer system, while in other embodiments multiple such systems, or multiple nodes making up computer system, may host different portions or instances of embodiments. For example, in one embodiment some elements may be implemented via one or more nodes of computer systemthat are distinct from those nodes implementing other elements.

In various embodiments, computer systemmay be a uniprocessor system including one processor, or a multiprocessor system including several processors(e.g., two, four, eight, or another suitable number). Processorsmay be any suitable processor capable of executing instructions. For example, in various embodiments, processorsmay be general-purpose or embedded processors implementing any of a variety of instruction set architectures (ISAs), such as the x86, PowerPC, SPARC, or MIPS ISAs, or any other suitable ISA. In multiprocessor systems, each of processorsmay commonly, but not necessarily, implement the same ISA.

In some embodiments, at least one processormay be a graphics processing unit. A graphics processing unit or GPU may be considered a dedicated graphics-rendering device for a personal computer, workstation, game console or other computing or electronic device. Modern GPUs may be very efficient at manipulating and displaying computer graphics, and their highly parallel structure may make them more effective than typical CPUs for a range of complex graphical algorithms. For example, a graphics processor may implement a number of graphics primitive operations in a way that makes executing them much faster than drawing directly to the screen with a host central processing unit (CPU). In various embodiments, graphics rendering may, at least in part, be implemented by program instructions that execute on one of, or parallel execution on two or more of, such GPUs. The GPU(s) may implement one or more application programmer interfaces (APIs) that permit programmers to invoke the functionality of the GPU(s). Suitable GPUs may be commercially available from vendors such as NVIDIA Corporation, ATI Technologies (AMD), and others.